Research map / Vision-language and multimodal / Visual question answering
Locate Then Generate: Bridging Vision and Language with Bounding Box for Scene-Text VQA
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| Add answer-region localization and language refinement to BART (ablation) | BART baseline · ST-VQA val | accuracy: 48.02 -> 52.21 | Helped |
| Add region-proposal visual word selection (ablation) | BART baseline · ST-VQA val | accuracy: 48.02 -> 50.52 | Helped |
| Add language-refinement word selection (ablation) | BART baseline · ST-VQA val | accuracy: 48.02 -> 50.48 | Helped |
| Replace separate visual and language selection with visual selection only (ablation) | LTG · ST-VQA val | accuracy: 52.21 -> 50.52 | Hurt |
| Replace separate visual and language selection with language selection only (ablation) | LTG · ST-VQA val | accuracy: 52.21 -> 50.48 | Hurt |
| Use BART instead of the prior best non-pre-training baseline (ablation) | ST-VQA · ST-VQA val | accuracy: 45.29 -> 48.02 | Helped |
| Use LTG instead of M4C | TextVQA · TextVQA val | accuracy: +6.06% | Helped |
| Use LTG instead of LOGOS | TextVQA · TextVQA val | accuracy: 50.79 vs 50.56 | Hurt |
| Jointly train LTG with extra TextVQA data | LTG · ST-VQA val | accuracy: +0.49% | Helped |
| Jointly train LTG with extra TextVQA data | LTG · ST-VQA val | ANLS: +0.3% | Helped |
| Use LTG rather than TAP | ST-VQA · ST-VQA val | accuracy: +6.92% | Helped |
| Use Microsoft-OCR rather than Rosetta-en with M4C | M4C · TextVQA val | accuracy: 39.40% -> 44.50% | Helped |
Closest papers by meaning
- FOCUS: Internal MLLM Representations for Efficient Fine-Grained Visual Question Answering (NeurIPS 2025 · similarity 0.90)
- Marten: Visual Question Answering with Mask Generation for Multi-modal Document Understanding (CVPR 2025 · similarity 0.90)
- Track the Answer: Extending TextVQA from Image to Video with Spatio-Temporal Clues (AAAI 2025 · similarity 0.90)
- When Open-Vocabulary Visual Question Answering Meets Causal Adapter: Benchmark and Approach (AAAI 2025 · similarity 0.89)
- BLIVA: A Simple Multimodal LLM for Better Handling of Text-Rich Visual Questions (AAAI 2024 · similarity 0.89)
- Cross-Lingual Text-Rich Visual Comprehension: An Information Theory Perspective (AAAI 2025 · similarity 0.89)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.